Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport

Mingyang Sun, Pengxiang Ding, Weinan Zhang, Donglin Wang
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:57374-57390, 2025.

Abstract

Diffusion policies have shown promise in learning complex behaviors from demonstrations, particularly for tasks requiring precise control and long-term planning. However, they face challenges in robustness when encountering distribution shifts. This paper explores improving diffusion-based imitation learning models through online interactions with the environment. We propose OTPR (Optimal Transport-guided score-based diffusion Policy for Reinforcement learning fine-tuning), a novel method that integrates diffusion policies with RL using optimal transport theory. OTPR leverages the Q-function as a transport cost and views the policy as an optimal transport map, enabling efficient and stable fine-tuning. Moreover, we introduce masked optimal transport to guide state-action matching using expert keypoints and a compatibility-based resampling strategy to enhance training stability. Experiments on three simulation tasks demonstrate OTPR’s superior performance and robustness compared to existing methods, especially in complex and sparse-reward environments. In sum, OTPR provides an effective framework for combining IL and RL, achieving versatile and reliable policy learning.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-sun25c, title = {Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport}, author = {Sun, Mingyang and Ding, Pengxiang and Zhang, Weinan and Wang, Donglin}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {57374--57390}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/sun25c/sun25c.pdf}, url = {https://proceedings.mlr.press/v267/sun25c.html}, abstract = {Diffusion policies have shown promise in learning complex behaviors from demonstrations, particularly for tasks requiring precise control and long-term planning. However, they face challenges in robustness when encountering distribution shifts. This paper explores improving diffusion-based imitation learning models through online interactions with the environment. We propose OTPR (Optimal Transport-guided score-based diffusion Policy for Reinforcement learning fine-tuning), a novel method that integrates diffusion policies with RL using optimal transport theory. OTPR leverages the Q-function as a transport cost and views the policy as an optimal transport map, enabling efficient and stable fine-tuning. Moreover, we introduce masked optimal transport to guide state-action matching using expert keypoints and a compatibility-based resampling strategy to enhance training stability. Experiments on three simulation tasks demonstrate OTPR’s superior performance and robustness compared to existing methods, especially in complex and sparse-reward environments. In sum, OTPR provides an effective framework for combining IL and RL, achieving versatile and reliable policy learning.} }
Endnote
%0 Conference Paper %T Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport %A Mingyang Sun %A Pengxiang Ding %A Weinan Zhang %A Donglin Wang %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-sun25c %I PMLR %P 57374--57390 %U https://proceedings.mlr.press/v267/sun25c.html %V 267 %X Diffusion policies have shown promise in learning complex behaviors from demonstrations, particularly for tasks requiring precise control and long-term planning. However, they face challenges in robustness when encountering distribution shifts. This paper explores improving diffusion-based imitation learning models through online interactions with the environment. We propose OTPR (Optimal Transport-guided score-based diffusion Policy for Reinforcement learning fine-tuning), a novel method that integrates diffusion policies with RL using optimal transport theory. OTPR leverages the Q-function as a transport cost and views the policy as an optimal transport map, enabling efficient and stable fine-tuning. Moreover, we introduce masked optimal transport to guide state-action matching using expert keypoints and a compatibility-based resampling strategy to enhance training stability. Experiments on three simulation tasks demonstrate OTPR’s superior performance and robustness compared to existing methods, especially in complex and sparse-reward environments. In sum, OTPR provides an effective framework for combining IL and RL, achieving versatile and reliable policy learning.
APA
Sun, M., Ding, P., Zhang, W. & Wang, D.. (2025). Score-Based Diffusion Policy Compatible with Reinforcement Learning via Optimal Transport. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:57374-57390 Available from https://proceedings.mlr.press/v267/sun25c.html.

Related Material